Natural and Artificial in the Language of the Malayalam Text

Yuktibhās.ā

P.P.Divakaran
Chennai Mathematical Institute, SIPCOT IT Park, Siruseri 603 103 (India)

1
I. INTRODUCTION

In 1832, Charles M. Whish, an Englishman who had worked for the East India Company,
presented to the Royal Asiatic Society of Great Britain and Ireland an account[1] of the
contents of four palm leaf manuscripts which he had found in the environs of Cochin (or
Kochi as it is now known) in coastal central Kerala. One of the four is a work in Malayalam,
the language of Kerala, on mathematics and astronomy entitled Yuktibhās.ā (abbreviated to
YB from now on; the others are also mathematical/astronomical texts, by now similarly well-
known, in Sanskrit). Thus was brought to the notice of scholarly Europe, perhaps for the
first time, the text of which I speak here. Whish’s effort seems to have made no impression
on historians of mathematics till the 1940s. But the decades since then have seen increasing
attention being paid[2] to the quite astonishingly sophisticated mathematics produced by
the socalled Kerala school in a relatively brief period of 200 years or so, beginning in the
second half of the 14th century CE.
The text of YB that I use here is the one that virtually everyone interested in the work
relies on directly or indirectly, that given in the annotated Malayalam edition of Rama
Varma (Maru) Tampuran and A. R. Akhilesvara Ayyar, published in 1948[3]. It is based
on four separate manuscripts which are largely in concordance. Tampuran belonged to a
local royal family and was a well-known scholar of Malayalam (and hence of Sanskrit) and
of the śāstras, and Ayyar was a schoolteacher with a master’s degree in mathematics; in
the complementarity of their domains of expertise, a more ideal pair of collaborators is hard
to imagine. The main part of the book is 290 printed pages long of which roughly half
is commentary. The editing and the commentary are impeccably done, in the best Indian
tradition of bhās.yas. Especially noteworthy are the meticulously drawn geometric figures
without which the work of those who followed them would have been made greatly more
arduous.
The edition of Tampuran and Ayyar covers only Part I of YB. Part II contains applications
to astronomy and has so far not been published, in any language. The late K. V. Sarma,
whose efforts more than of anyone else brought the main texts of Kerala mathematics and
astronomy to the attention of the scholarly world, had completed (in association with M.
D. Srinivas, M. S. Sriram and K. Ramasubramanian) an English traslation of both parts of
YB at the time of his death in January 2005, but it is yet to appear. As of now, a faithful

2
presentation of what YB contains, in any language other than Malayalam (unfortunately
beyond the acquaintance of most interested scholars) does not exist.

II. THE BACKGROUND

Till about 700 CE or so, Kerala formed part of the Tamil linguistic landscape. The
rich cultural, especially literary, life of those days is well documented and we also have
some knowledge of the polity. There is however no serious evidence of any astronomical or
mathematical activity at that time in Tamil Kerala nor indeed in the Tamil country as a
whole. Very plausibly, the taste for astronomy and mathematics travelled to Kerala with a
series of waves of migration from further north, generally along the coast, of vedic brahmins.
Beginning in the 7th – 8th century CE, this influx continued for half a millennium or longer.
The early settlers bestowed Aryan legitimacy on local potentates who built temples for them
in return, with extensive land grants for the maintenace both of the temples and, in some
comfort, of the temple managers. But these brahmins, who came to be known generically as
namputiris, brought with them a great deal more than vedic expertise; they were, at least
some of them, a learned people, the repositories and practitioners, creators and transmitters
of all the classical sciences. Thus it is that they carried to Kerala a strong heritage in
mathematics and astronomy, most notably the legacy of Aryabhata. They also brought
with them their language of learning, Sanskrit.
But largely independent of northern influences, the old Tamil language of the region had
already begun to take on a distinct local personality. On top of this came the impact of
Sanskrit, resulting in a rich and versatile new language, Malayalam. By the time of the
composition of YB, Malayalam had settled down to its definitive form, hardly different from
what it is today. Nevertheless, with the exception of YB and a few minor texts, Sanskrit
continued as the language of mathematical scholarship (and writing) until its final decline
and disappearance, as sudden as was its beginning[4].
A number of literary references and inscriptions from this period of transition testify
to the existence all over Kerala of educational institutions, ranging from the veda schools
attached to temples to veritable quasi-autonomous colleges. Over the first half of the sec-
ond millennium CE, these schools, together with the academies that flourished under royal
patronage, were an integral part of the intellectual vibrancy of the time. It is in this setting

3
that almost all of the literally thousands of scientific manuscripts now lying scattered in
private and public collections came to be composed[5].
The earliest known mathematical work which can be unambiguously attributed to Ker-
ala actually predates this golden age. This is a Sanskrit verse commentary by Shankara-
narayanan[6] on the first Bhaskara’s Laghubhāskarīyam. According to the text itself,
Shankaranarayanan was the royal astronomer to the Cera emperor Sthanu Ravi Varma
– the Ceras had lately consolidated their control over most of Kerala – working in the ob-
servatory in the capital Mahodayapuram (the famed Muziris or Musiris, emporium to the
world; modern Kodungallur). The text gives the date of composition as 869 CE and refers
to the author as hailing from Kollapuri in Paidhyarashtra (Pratishthana or Paithan) which
has been identified (see note [5]) very persuasively with modern Kolhapur in Maharashtra.
This is rich and rare fare: in one text, we have a precise date and place of composition and
a precise provenance for the author or his immediate ancestors.
As far as we know, the emergence of astronomical scholarship in Kerala with Shankara-
narayanan was an isolated event. His intellectual ancestry is unknown and no one seems to
have carried on after him; in fact there is no sign of any serious mathematical and astro-
nomical activity in Kerala in the next five hundred years. To an extent, this can be blamed
on the war and turmoil that visited Kerala from the Tamil country at the turn of the first
millennium. By the end of the 11th century, the prosperous Cera kingdom was gone forever,
its capital razed, the royal observatory only a memory.

III. MADHAVAN AND HIS FOLLOWERS

In the chaotic aftermath of this ‘100 year war’, the Cera kingdom broke up into a number
of principalities, of which the most powerful was that ruled by the Zamorin (samutiri)
dynasty from Calicut. This period also saw a renewed influx of brahmins and their rise to
a position of social and cultural preeminence. The southern part of the Zamorin’s domain
became a particularly favoured area for namputiri settlements. It is in this region, on
either bank of the river Nila (more commonly known now as the river Bharata), that there
emerged, beginning in the second half of the 14th century, a succession of astronomers and
mathematicians of truly exceptional quality. They formed a tight-knit group, all linked
together by the traditional teacher-disciple bonds. There was at least one father-son pair

4
among them, Paramesvaran and Damodaran. Most were namputiris of one variey or another.
According to local tradition, three were native to one village and two taught at the same
temple school. Virtually all lived in close geographic proximity and produced their work
within a time span of roughly two centuries.
The most original and creative of this brilliant lot, by the acclamation of those who
followed him, was also the first: Madhavan, said to be of Sangamagramam[7]. Madhavan
has remained a somewhat shadowy figure. We know nothing about his life except that
it spanned the second half of the 14th century and the first quarter of the 15th and that
the great astronomer Paramesvaran was his disciple. The few manuscript fragments that
are attributed to him are primarily astronomical, having little mathematical content of any
originality, and it is through the writings of his intellectual heirs that his achievements
are known to us. Paramesvaran’s student Nilakanthan, in particular, quotes mathematical
verses attributed to Madhavan quite liberally in his Sanskrit work Tantrasam
. graha, making
it clear that it is meant to be a compendium of his teaching. The overwhelming significance
of YB lies in the fact that it contains, in chapters 6 and 7, the most thorough exposition
we have of the new mathematics that Madhavan created and Tantrasam
. graha summarised.
If only to place YB itself in context, it is useful to touch on the highlights of Madhavan’s
legacy.
Since before the time of Aryabhata, the one abiding theme of Indian mathematics was
the circle, more particularly the relationship between an arc and the corresponding chord,
and a variety of questions linked to it. This was entirely in line with the uses to which
mathematical reasoning was put in India, first and foremost for the study of the geometry of
the celestial sphere and the motion of heavenly bodies in their epicyclical orbits. Aryabhata
himself famously recognised the difficulty of giving an exact number for the ratio of the
circumference of a circle to its diameter, qualifying his value 3.1416 for π as approximate,
āsanna. Fundamentally, what Madhavan did was to push the quest for precise values for
such ratios to its logical and mathematical limit. (That π was understood by this time to
be an irrational number is clear from a well-known passage from Nilakanthan’s commentary
on the Āryabhat.īya, see below). In the process, he was led, at a technical level, to express
such ratios (trigonometric functions) as appropriate power series, infinite series of smaller
and smaller terms with none exactly equal to zero. (We know them as the ‘Gregory-Leibniz’
series for the arctangent and the ‘Newton’ series for the sine and the cosine, the quotation

5
marks indicating a compromise between current usage and historical accuracy. I shall write
these series down as the occasion arises). Madhavan also did some wonderful and practical
things with these series: techniques for accelerating their convergence; accurate estimates
of the remainder as rational functions when a series is truncated; interpolation formulae for
values of trigonometric functions at a point in terms of the values at a neighbouring point;
and so on. Most of this is described in detail in YB, the interpolation formulae being a
notable exception.
But more profoundly, Madhavan arrived at a general method, a philosophy almost, of
addressing such questions by passing to the infinitesimally small and then summing the
infinite number of the resulting infinitesimal contributions to the relevant geometrical quan-
tity. Stated simply, Madhavan invented calculus, as it applies to circular arcs. Subject only
to the limitation to circular arcs, YB conveys clearly that the key conceptual step in this
was the recognition that local approximation by linear functions (tangents), in other words
differentiation, and their subsequent summing up, integration, are converse processes – in
essence, the earliest version of what came to be known in its sharp subsequent form as the
fundamental theorem of calculus. What is most striking in all this to the present day reader
brought up in the mathematical culture of the 19th and 20th centuries is the easy mastery
with which the supposed twin demons of the infinite and the infinitesimal are simultaneously
tamed.

IV. ABOUT YB

K. V. Sarma’s date for the composition of YB, around 1550-1560 CE, seems now to be
generally accepted, as also his attribution of the authorship to Jyeshthadevan[8]. We know a
little more about Jyeshthadevan than about Madhavan: he was a disciple of Nilakanthan and
Damodaran, both disciples of Madhavan’s disciple Paramesvaran and, like Paramesvaran,
was connected with the Rama temple (or to a school attached to it) in the village of Alattiyur,
just north of the river Nila not far from its mouth. And, most importantly, we have a whole
book, extensive and intact, attributed to him[9]. Alattiyur still has memories of his name
and of its past mathematical-astronomical glories, but astrology has long since supplanted
astronomy as its chief source of pride.
In the context of Indian mathematical writing, YB is singular in several respects: i) it is

6
not in Sanskrit verse but in Malayalam (bhās.ā) prose, ii) it provides detailed lines of reasoning
(yukti) and, as a consequence, iii) it is inordinately long. It is also very ‘theoretical’; there
are few worked examples or numerical illustrations (udāharan.ams) unlike say in the work
of the second Bhaskara and, despite the strongly geometric methods of proof, no diagrams
at all. (Tampuran and Ayyar have included a profusion of them in their commentary and
these have been freely and gratefully duplicated in the subsequent literature). It is relevant
to ask why it is three-fold singular in this particular way. Given our ignorance of the
circumstances of its writing, the answers can only be tentative at best. But, reading the
text, it is difficult to escape the conviction that, unlike the traditionally favoured format
of sūtras, mnemonic verses serving as an aide-mémoire with the reasoning being explained
in face-to-face sessions, YB was meant to be autonomous, written down from the guru’s
mouth, to be read, struggled with and, hopefully, mastered in course of time, away from
the guru and the classroom. The traditional compact verse format in mathematical writing
had at least one function other than that of facilitating (primarily oral) communication
and text-preservation; concise expressions are commonly made to stand in for numbers,
formulae and even whole collections of ideas and methods, ‘packages’ to be unzipped in the
mind. YB does have such packages scattered throughout, of its own and in the form of
quoted verses from other sources, but on the whole stands at the opposite extreme. The
entire text is in a natural, too natural perhaps, Malayalam, explicit and even verbose, with
the Sanskrit technical terms embedded in this matrix. The style is colloquial and often
repetitive, especially in its use of connecting words common in conversation but not serving
any functional (let alone mathematical) purpose. Altogether, its writing was a startlingly
original endeavour. One cannot escape the feeling of reading the faithful notes of a masterly
but informal set of lectures – face-to-face presentations – but without the easy present day
option of an escape to equations and diagrams[10].
YB claims no originality of content; the first sentence is: “In order to explain[11]
all the mathematics useful for the motion of heavenly bodies (grahas), according to
Tantrasam
. graha[12], I begin by describing the general (or common) mathematical opera-
tions such as addition and so on”. Three short chapters deal routinely with these basic
operations. An even shorter (considering the heavy use made later of the properties of sim-
ilar triangles) chapter 4 concerns the ‘rule of three’ (trairāśikam). A substantial chapter
5 (together with a long appendix, explicitly stated to be taken over from Tantrasam
. graha,

7
with several numerical examples) is devoted to the method of ‘pulverisation’ (kut..tākāram)
for the Diophantine solution of an equation of the first degree in two unknowns. This will
find its use in Part II of YB. As already noted, the meat of the work is in the last two
chapters, 6 (the relationship between the circumference and the diameter) and 7 (the theory
of chords). Including the commentary, they run to over 200 printed pages.
The inevitable question that arises is how effective the use of an almost totally conversa-
tional Malayalam is in conveying the subtle and complex reasoning that this new knowledge
is based on and, more generally, what obstacles the reliance on such a natural style of
thought and communication places in the way of development of what is after all a highly
abstract undertaking. To enable readers to judge for themselves how well these challenges
are met, I now present in translation two passages from YB, both relating to calculus, before
expressing my own views. The first describes the computation of the surface area of a sphere
(from chapter 7) and the second is the section in which the integrals of the positive integral
powers of a variable are worked out (in chapter 6).
The translation is as literal as I can sensibly make it. The only liberties I have taken are
to provide basic punctuation and paragraph breaks, to drop the too-frequent conversational
connectives, to supply personal pronouns where English requires them and to add a few
brief explanations (not present in the original) or comments [within square brackets]. In
particular, I provide no diagrammatic guide to the reasoning (the basic figure which I use
to define notation for my own explanation of the passage on integration of powers should
not count) – those so inclined will find it instructive, at least as a start, to supply their own
figures before looking up Tampuran and Ayyar[3] or Sarasvati Amma[2]. The first passage
requires only a minor transformation to make it ‘modern’; Sarasvati Amma’s version is
largely faithful. The second passage holds a few points of mathematical interest which it
will be worthwhile to return to later. It is also the part of Kerala calculus that today’s
scholars, unlike Tampuran and Ayyar, sometimes tend to handle with less than complete
fidelity, lapsing occasionally into a reliance on later western developments like the binomial
theorem for negative exponents, theorems on limits and their interchangeability with infinite
sums etc., in place of what is actually in YB. For this reason, in the introductory remarks
on this passage and in the few explanations within it, I have made an effort to guard against
my use of current notation distorting in any way the material I am trying to convey.

8
V. THE SURFACE AREA OF A SPHERE

The section headed golapr..s.thaks.etraphalānayanam gives the proof of the formula: surface
area of a sphere = diameter × circumference [of a great circle]. It occurs at the very end
of chapter 7 (with only the calculation of the volume of the sphere to follow) though, with
the exception of the use of one ‘package’, the work required is much less demanding than
in the proofs of the various trigonometric series earlier in the book. It is probably the case
that Madhavan invented the infinitesimal method first for the determination of π via the
‘Gregory-Leibniz’ series and then found it to be a powerful general tool for settling several
other interesting problems[13] (also see below, section I).
Here is the translation.

Now I narrate that, combining two principles just explained, [namely that] from
pin.d.ajyāyogam can be produced khan.d.āntarayogam and [that] knowing the di-
ameter at one place, we can apply the rule of three (do a trairāśikam) as we
please, the area of the surface of a sphere will arise.

A uniformly rounded object is called a sphere (golam). Through the middle

of such a sphere, imagine two circles, one along east-west and the other along
south-north. Then imagine circles, one shifted slightly to the south and the
other slightly to the north of the east-west circle [the equator]. Their distances
from the east-west circle should be the same for all parts (avayavam) [the word
samāntaram for parallel is not used in this section though it is, for straight lines,
elsewhere]. Consequently, these two will be slightly smaller than the first (or
original) one. Then, starting from these, imagine slightly smaller and smaller
circles, all of them at equal distance one from another [i.e., between successive
latitudes], so as to end at the south and north edges [the poles]. Their separation
along the south-north circle must be seen to be equal. This being so, imagine
that the circle-shaped gap between two circles [successive latitudes] is cut at
one place, removed and straightened (or spread). Then, of the circles on the
two sides of the gap, the larger one will be the base (bhūmi) and the smaller
one the opposite side (mukham) of a trapezium (samalam
. bacaturaśram) whose
lateral sides (pārśvabhuja) will be the separation (antarālam) along the south-
north circle of [two successive latitude] circles. Now cut out the part outside

9
the altitude (lam
. bam) [from an upper vertex to the base], turn it upside down
and transfer it to the other side [the opposite edge]. This is a rectangle whose
length is half the sum of the base and the oposite edge and whose width is the
altitude. In this way, think of all the gaps (antarālam) [elsewhere a non-Sanskrit
Malayalam word is also used] as rectangles (āyatacaturaśram). Their widths
are all equal. Lengths have various (or varying) measures (pramān.am). The
result of multiplying the length and the width is the area (ks.etraphalam). The
widths of all being equal, add the lengths of all and multiply by the width. Thus
will arise the area of the surface of the sphere. [Nothing so far about the passage
to the limit].

Next, the method (upāyam) to know how many gaps there are and what their
lengths and width are. The radii of the circles that we have imagined above [the
latitudes] are half-chords (ardhajyā) of a circle whose radius is the radius of the
sphere. Hence, multiplying these half-chords by the circumference of the sphere
and dividing by the radius of the sphere [i.e., multiplying by 2π] will result in
circles [their circumferences to be precise] having the half-chords as radii. These
will be the lengths of the rectangles if the chords are taken at the midpoints of
the gaps. Multiplying the sum of all half-chords (ardhajyāyogam) [by 2π, left
unsaid] will result in the total length of all the figures (ks.etrāyāmam) [i.e., the
sum of the lengths of all the rectangles]. The gap between two arbitrary [the
Malayalam phrase is yāva cilava = yāvat tāvat] circles [an arbitrarily chosen pair
of consecutive latitudes is meant] at the south-north circle mentioned earlier is
an arc segment (cāpakhan.d.am) of the circumference of the sphere [i.e., of the
S-N circle].

Next, the method of getting the sum of the chords (jyāyogam). Multiply the
square of the radius by khandāntarayogam [I make no attempt at an exact
translation of this ‘package’. The expression stands for the sum of the sec-
ond differentials of the half-chords with respect to the arc (or, equivalently, the
angle) and has already been treated earlier in the chapter in connection with
the series expansions of sine and cosine. The result about to be used below is
also derived there.] and divide by the square of the full chord of the arc seg-

10
 ment (cāpakhan.d.asamastajyāvarggam). The result is the sum of the half-chords.
Multiply this by the width. The width is the chord of the arc segment. The
khan.d.āntarayogam is the first chord segment. Because of smallness [of the arc
segment] these [the first chord segment and the width] are [both] almost equal
to the full chord. These are the multipliers and the square of the full chord is
the divisor. But multiplication and division are unnecessary [the square of the
full chord cancels from the numerator and denominator]. What remains is the
radius. This has to be multiplied by the circumference and divided by the radius.
Only the radius will survive. Since we have to get the result for both halves of
the sphere, double the radius. Therefore, multiplying the diameter of the sphere
by the circumference of the sphere will produce the area of the surface of the
sphere.

We have here a straightforward narration in natural spoken Malayalam, understandable

literally by any literate person but for the technical terms. By and large, the technical
nomenclature is that used in mainstream mathematical writing in Sanskrit over many cen-
turies and is consequently unambiguous. But these linguistic resources are now asked to be
at the service of a sophisticated novel mode of reasoning demanded by Madhavan’s infinites-
imal mathematics, a service they were not designed to provide in full. Added to this burden
is the yoke of a foreign tongue, Malayalam, which was not sung, at least mathematically
speaking, at the cradle of the author. The strain shows. Even in translation it is evi-
dent that, while the first part of the proof (the geometric construction) is simply presented
and easy to follow, the transition to the calculus part is marked by an increasing opacity.
What the translation does not show is that the Malayalam itself is, apart from being very
informally used, quite unrefined in relation to other literary material of the period.

VI. INTEGRALS OF POWERS

Chapter 6 of YB is devoted to establishing the validity of the ‘Gregory-Leibniz’ series,

first for the angle π/4 and then for general angles, and to various improvements on it. Apart
from the mathematical content itself, it is noteworthy as the first illustration of the power of
the new infinitesimal geometry when combined with the equally new technique of integration
or ‘summation in the limit’, sam
. kalitam.

11
 The geometry considers a circle of radius r and a square of side 2r in which the circle is
inscribed, focussing on one octant of the circle as in the diagram below:
A P B

The side AB is divided into n equal segments of length d (P is a typical point of this
division), with the limit d → 0, n → ∞ (with nd = r fixed) to be taken subsequently. After
a series of steps including the usual clever choices of similar triangles (trairāśikam), some
simple algebraic identities, a process of iteration and a judicious neglect of terms of order
1/n2 (i.e., second order infinitesimals in the limit), the geometry is shown to lead to the
result
π Xn
1 j2 j4
arclength of the octant = r = r lim ( − 3 + 5 − ...),
4 j=1
n n n

lim meaning the limit described above. Each term within the bracket (including the sum
over j and in the limit) is referred to as a sam
. kalitam which I shall translate as an integral,
in anticipation[14]. In addition to the even sam
. kalitams contributing to the arc length, YB
considers also the odd sam
. kalitams since the induction procedure requires their evaluation.
Thus, ignoring the minus signs, the kth sam
. kalitam is defined to be
Xn
rj k
Ik := lim , k = 0, 1, 2, ...
j=1
nk+1

The introductory paragraph of the section headed sam

. kalitams is:

Here I describe the method of producing the integrals. First the simple integral
(kevalasam
. kalitam) is described. Then the integral of two equals multiplied
together. Then, even though it is not useful here, I describe also integrals of
equals multiplied by themselves three, five, etc. times, since they occur in the
midst of those which are useful [namely, the even powers].

12
The term kevalasam
. kalitam does not reappear in the section. In this passage it evidently
means the integral of the first power, which is referred to in the rest of the section as
mūlasam
. kalitam. The distinction is with the sequence of repeated k-fold integrals of x
(ādyadvitīyādisam
. kalitams); i.e., in modern notation,
Z Z Z
xk+1
dx dx... dx =
(k + 1)!

which are taken up further on to generate the factorial denominators in the ‘Newton’ series.
With this potential confusion out of the way, here is the section entitled mūlasam
. kalitam on
the computation of the integral

r 2r (n − 1)r r d
I1 = lim( + 2 + ... + + ) = lim (d + 2d + ... + (n − 1)d + r),
n 2 n n 2 n r

the second form being the expression with which YB actually works.

In mūlasam
. kalitam, the last side (bhujā) [a bhujā is the side AP of the right
triangle OAP , OP being the corresponding karn.n.am; so the last bhujā means
AB] is equal to the radius of the circle, the one below (or before) that is one
segment (khan.d.am) less and the one before that two segments less. Suppose all
the sides are equal to the radius. In that case, if the radius is multiplied by
the number of sides [i.e., n], that will be the result of the sam
. kalitam. But here
only one side is equal to the radius. Starting from this, the sides of the other
smaller and smaller diagonals (karn.n.ams) are, in order, one unit (sam
. khya) at
a time less. Whatever is the number of units the radius is supposed to have,
imagine that the number of segments of the side [here it means the full side AB]
is the same. That makes it easy to remember. The last but one side will be one
unit less. The next shorter one will be two less than the number of units of the
radius. The missing part (am
. śam), starting with one, will increase one [unit] by
one [unit] progressively, the last deficit (ūnāmśam) almost equal to the radius,
just one unit less. Now if the deficits are all added, this number (sam
. khya) will
eventually equal the sum [the word used for this finite sum is sam
. kalitam, as
is traditional, see note [14]] of terms starting with one, increasing by one and
ending with the radius, less one radius [i.e., it is the original sum minus the
radius]. Therefore, multiply the number of units in the radius by one added to

13
 the number of sides [these two numbers are the same by assumption]. Its half
is the bhujāsam
. kalitam. Bhujāsam
. kalitam means the sum of the sides of all the
diagonals.

The smaller the segment, the more accurate (sūks.mam) the result. Therefore
imagine that each segment (bhujākhand
. am) [the text has bhujāsam
. khya, which is
probably a result of sloppiness somewhere in the transcription] is divided (cut)
into atoms (an.u) and then carry out the sam
. kalitam. For this, if the division
is by parārddham [any very large number will do; here, specifically, it is 1017 ],
add one to the product of the bhujā and parārddham, multiply by the radius and
halve it. Also, divide by parārddham. This is approximately half the square of
the radius. To make it a whole number, divide by parārddham. Thus, to the
extent that the segment is short, only a small part has to be added to the bhujā to
produce the sam
. kalitam. [I understand this sentence to mean that the correction
to the integral due to the finite length of the segment can be neglected in the
limit]. Therefore, adding nothing to the bhujā, multiplying it by the radius and
halving it will result in the sam
. kalitam of the extremely finely segmented bhujā.
This is how half the square of the radius is the accurate bhujākhan.d.asam
. kalitam.

The terminological imprecisions in the passage (for instance the different senses in which
the words bhujā, sam
. khya and even sam
. kalitam itself are employed) are evident as is the
increasing lack of clarity (the parārddham business for example), once again, of the writing
when it comes to describing the reasoning about the passage to the limit.
After this account of the computation of the integral of x, the next subsection deals
with x2 in some detail. The passage from x to x2 involves conceptual and technical
novelties which then carry over smoothly to the general inductive step, from xk to xk+1 .
Here is how the general method of computing the integral of an arbitrary positive power
(sam
. kalitānayanasāmānyannyāyam) is summarised:

....To make integrals of higher and higher powers, multiply the particular integral
by the radius and remove from it the result divided by the number which is one
greater [Ik+1 = xIk − xIk /(k + 1) in our notation]. Thus divide the square of the
radius by two, the cube by three, the fourth power by four, the fifth power by

14
 five. Thus divide the consecutive powers starting with the first (ekaikottarasam-
aghātam) by [the same] consecutive numbers; the results will be the integrals of
powers in increasing order. As the simple integral comes from the square, the
integral of the square from the cube, the integral of the cube from the fourth
power and so on, the power of an unknown (rāśi) when divided by the same
number as the power will give the integral of one less power of the unknown.
This is the method of producing integrals of all powers.

There are several mathematically interesting reasons for reproducing this brief paragraph,
to which I will return shortly. But for the present, let us overlook the familiar prolixity
(perhaps it was a class of dull students) and note the clarity with which the theorem itself
is stated.

VII. THE LANGUAGE

It bears repetition that, as in the excerpted passages, so in the entire text, no symbols are
employed to represent the mathematical objects being manipulated, no formal notation for
relations among them and operations on them, no diagrammatic guide to the geometric con-
structions invoked. The only devices which can be considered as perhaps artificial are, first,
the use of common natural (generally Sanskrit-origin) words to denote precisely understood
geometric entities: examples include cāpam, jyā, śaram, bhūmi, bhujā, mukham, etc; sev-
eral of these and other similar terms, when translated, have exactly the same significance in
European geometry. (The use of natural words as units of an artificial language appropriate
for a given discipline is of course a widespread and contemporary practice even in extremely
abstract contexts, see the current literature in mathematics and physics). Then there are
the phrases whose sense is generally clear from the way they are formed as prescribed in
the usual rules of compounding in Sanskrit (and Malayalam), e.g., samalambacaturaśram,
vyāsārddhavarggam, etc. – they have no extra or hidden meaning apart from the literal.
Finally, beyond these, there occur certain expressions whose literal meanings are inadequate
to convey exactly what they represent or may even be misleading. The most widespread
instance of this in YB is trairāśikam, ‘the rule of three’, encompassing all properties of a set
of four numbers in proportion, and hence all properties of a pair of similar triangles. In the
passage on the area of the sphere, the term pin.d.ajyāyogam is also such a ‘package’, standing

15
for a certain refined procedure which effectively computes an integral: it means the sum of
semichords of arcs of equal length into which a quadrant of a circle is divided, in the limit
of vanishing arc length. The expression khan.d.āntarayogam stands for an even more subtle
procedure – bhujākhan.d.am corresponds to the differential and [bhujā]khan.d.āntaram to the
second differential of the sine of an angle. (Thus the introductory sentence of the passage
on the surface area of the sphere is equivalent to the statement that the second differential
of sine is proportional to itself). Probably the best known instance in Kerala mathematics
of such esoteric usage is the phrase jīveparasparam attributed (in Tantrasam
. graha) to Mad-
havan and representing the formulae for the sines of the sum and difference of two angles
(together with the proof). These are used in YB as part of the preparation for the con-
struction of the ‘Newton’ series for sine and cosine. Unwrapping such ‘packages’ is essential
for the understanding of the yukti, but they circumscribe the maximum deviation from the
natural use of language that can be found in YB.
One other point about the language, or rather the general style of discourse, merits men-
tion. The second section of chapter 7 of YB has the heading ‘technical terms and definitions’.
A demanding reader looking here for precise characterisations of the concepts and terminol-
ogy necessary for the formulation and proof of the great theorem on the ‘Newton’ series to
follow is likely to be disappointed. The section is neither systematic nor complete, but only
a partial account of some of the geometric constructions and the associated nomenclature
needed for the proof. In all of YB, the terminology itself does not have an invariable signif-
icance. To give just one instance, the crucially important word sam
. kalitam is used to mean
both a finite sum as in earlier writing[14] and, in chapter 6, integrals of several different
types. In the same chapter 6, the term yogam never means an integral, but only a finite sum
whereas in chapter 7 we have, for example, arddhajyāyogam denoting the finite sum as well
as the limiting integral. On the other hand, sam
. kalitam is never used for the integrals oc-
curring in surface and volume computations and, elsewhere in chapter 7, khan.d.āntarayogam
and khan.d.āntarasam
. kalitam are distinguished.
Such instances can be multiplied. The question is: Is this lack of precision related in some
intrinsic way to a reluctance or an inability to transcend the limitations of natural means of
communication? Following from this: To what extent has the use of a dominantly natural
language impeded the processes of conceptualising and developing the mathematics and then
communicating it? The answer to the first question requires a thorough acquaintance with

16
texts contemporary with and anterior to YB (all written in the not very natural language of
terse mathematical Sanskrit) and is not for me to attempt. As for how well the mathematics
was conveyed, if Tampuran and Ayyar could read the work four centuries after it was written
and still make perfectly good sense of it – though they confess to having sometimes to guess
at the author’s train of thought – Jyeshthadevan clearly has not made too bad a job of it.

VIII. THE MATHEMATICS

What, now, are the mathematical ideas and methods, not just the unexpectedly novel
results and the clever computations, that are conveyed to us by the imperfect vehicle that is
the natural Malayalam of YB? The order in which the various series occur in YB – first the
numerical series for π/4, then the ‘Gregory’ series for general angle and then, in a veritable
tour de force, the two ‘Newton’ series for sine and cosine – would seem not to be a matter
of chance. It is a logical way to proceed if we accept that the motivation behind the search
for a series of ever smaller terms whose sum approaches, in the limit, the value π/4:

π 1 1
= 1 − + − ....
4 3 5

has probably to be sought in the conviction, going back to Aryabhata, that π is an irrational
number and therefore cannot be written as the sum of a finite number of fractions. Nilakan-
than, Madhavan’s true mathematical heir and Jyeshthadevan’s teacher, has this comment
62832
(in his Āryabht.īyabhās.yam) explaining why Aryabhata said the value π = 20000
was only
āsanna and, along the way, defining an irrational number. In loose translation: “Why is
an approximate value given here rather than the true (vāstava) one? Because it cannot be
expressed. A measure which measures the diameter without a remainder cannot measure
the circumference without a remainder and vice versa. We can only ensure the smallness of
the remainder, not its absence”. The lack of ‘remainderlessness’ is naturally accommodated
by an infinite series expansion (though of course every infinite series of fractions does not
sum up to an irrational number, as Nilakanthan well knew, cf. his remarks on convergent
geometric series in the same text), in a manner not very different from the way in which
infinite continued fractions arise in the use of the Euclidean algorithm.
As described in YB, the generalisation from an infinite series representation for the num-

17
ber π/4 to a power series expansion for an arbitrary angle θ(θ ≤ π/4) in terms of tan θ:

tan3 θ tan5 θ
θ = tan θ − + − ...
3 5

involves, technically, only a small step, an elementary application of trairāśikam. It has

no computational difficulty or significance and helps only marginally in determining a more
√
accurate value for π – replacing π/4 by π/6 (tan π/6 = 1/ 3), for example, improves the
convergence slightly. But conceptually it is a giant leap, leaving behind the prop of the
irrationality question and towards a notion of what we call today a function. It can be
credibly argued that it is this step and the series expansions of sin θ and cos θ in powers
of θ that together mark the advent of calculus, not just as a technique of calculation as
in the expansion of π/4, but as the beginning of a new discipline of analysis. The basic
operations of calculus make their appearance as needed in the course of these developments:
differentials of the first and second order (which are all that is required for the sine and cosine
functions), definite and indefinite integrals, the rule for integration by parts (in computing
recursively the integral of a general power), the notion of repeated integrals, the solution of an
elementary differential equation, namely d2 y/dx2 + y = 0 for the sine and cosine functions,
etc. Though the motivating impulses may not have been the same, it is uncanny to see
infinite series (‘equations with an infinite number of terms’ in Newton’s language) playing
such a decisive part alongside the basic concepts of differentials and integrals (‘fluxions’ and
‘fluents’) also in the early evolution of European calculus and analysis.
The assurance with which YB handles these power series should not really be surprising
once we recognise their close affinity with the recursive methods used systematically in
diverse areas of Indian thought, including earlier mathematical work. The most direct
parallel is perhaps with the positional notation for, say, positive integers. In his talk at
this Workshop on the positional ‘language’ for the writing of numbers, John Kadvany[15]
highlighted the fact that the positional representation of any positive integer is no more
than an abbreviation for a polynomial whose value it is, with coefficients from a finite set of
non-negative integers (0 to 9 in the decimal base case) when the variable is fixed at a positive
integer (= 10 in the decimal case) – as Kadvany stresses, what makes this representation
possible and powerful is the fact that integers can be added and multiplied to get other
integers. A set whose elements can be added and subtracted as well as multiplied (but not
necessarily divided) subject to the usual rules of addition and multiplication is an example

18
of a ring; the integers, both positive and negative, form a ring of a particularly simple kind.
So do polynomials, in the present case in one variable, and this common abstract structure
is the key to the positional notation. (When the variable is fixed at an integer, say 10, the
uniqueness of the positional representation is assured only if the coefficients are restricted to
be less than 10 and the usual carry over rules are imposed; an exact correspondence between
integers and their positional representatives requires this to be factored in). Arbitrarily large
integers are so representable by polynomials of arbitrarily large degree; there is no upper
bound on the integers that can be positionally represented.
The transition from the positional representation of unboundedly large integers to the
trigonometric power series now appears natural: substitute for the value of the base used
(10 for decimal) the appropriate variable (tan θ for ‘Gregory’, θ for ‘Newton’) and allow the
coefficients to take values in certain determinate fractions, both positive and negative. The
set of formal (i.e., ignoring possible nonconvergence) power series with coefficients which are
fractions do form a ring just as polynomials do.
Related to this is the point that the coefficients of a power series obviously cannot all be
enumerated. They can be specified by, and effectively only by, recursive rules in some variant
or another. In YB this is done in a number of slightly different but equivalent ways. In the
statement of the ‘Gregory’ series expansion, the first three coefficients are numerically given
and then one is instructed to take the general coefficients as the reciprocals of consecutive
odd integers with the signs alternating. For the ‘Newton’ series:
θ3 θ5
sin θ = θ − + − ....
3! 5!
θ2 θ4
cos θ = 1 − + − ...,
2! 4!
the recursive specification of the coefficients is more explicit. The coefficient s2k+1 of θ 2k+1
in the sine series is to be computed as
s2k−1
s2k+1 = .
(2k)2 + 2k
Together with the value s1 = 1, also specified, repeated use of this recursion formula leads
to
s2k−1 s2k−3 1
s2k+1 = = = ... = .
2k(2k + 1) (2k − 2)(2k − 1)2k(2k + 1) 1.2...(2k + 1)
A similar rule is given for the cosine series. It should be noted that the use of recursion is
not limited to just the statement of these formulae in an economical form. It is an organic

19
part of the development; the formulae for the coefficients introduce themselves recursively,
so to speak, in the very process of their computation. There is, moreover, a unity in the
way the coefficients in all these series arise. In the ‘Gregory’ series they are definite integrals
(from 0 to 1 or 0 to tan θ) of even powers and in the two ‘Newton’ series they are iterated
integrals of even and odd powers respectively, with positive and negative signs alternating.
It is accurate to say that the constructive proof of their correctness is itself recursive (though
I have to add that the proof is one of the least transparent passages in all of YB).
Perhaps the best illustration from YB of the use of recursion to establish the validity of
an infinite sequence of propositions is the short passage on the calculation of the general
sam
. kalitam given in translation earlier (section VI) together with the two sections of YB
which precede it. The untranslated (for reasons of space) sections give a recursive proof
for the correctness of the values of the integrals of powers. First I2 is computed in terms
of I1 , then I3 in terms of I2 , both in detail. This is followed by the statement that I4 can
be computed from I3 and, generally, Ik+1 from Ik in the same way. It is perfectly obvious
that the recursive method cannot work if all values of k, even and odd, are not considered
together, even though the series for π/4 requires Ik to be determined only for even k. Once
the first step in the recursion, I2 from I1 , is done, generalisation is straighforward, involving
as the common basic step a recognisable finite form of integration by parts. Indeed, a hint
that the recursive nature of the computation was fully understood is available in the choice
of the term mūlasam
. kalitam for I1 , the root sam
. kalitam from which all others are generated.
How close does all this come to the method of mathematical induction as it is understood
today? A modern formulation of the principle of induction, essentialy deductive, will go
something like: The truth of an infinite sequence of propositions Pk , k = 1, 2, ... is established
if i) P1 is true and if ii) Pk is true implies Pk+1 is true for all k. Lacking the necessary symbolic
aptitude, Madhavan and his followers could not possibly have expressed their computation-
oriented reasoning in such abstract and elegant language. But for one not versed in logical
niceties, it is difficult to see that their use of recursion, not only to generate an infinite
sequence of true mathematical statements but also to prove them, is in any fundamental
way different, once allowance is made for the general preference in India for constructive as
opposed to deductive proofs.
From a modern perspective, there are of course other issues raised by YB’s handling
of infinite series, chief among them the question of convergence. Not surprisingly, YB is

20
not concerned in any serious way with this question. It is striking however that all the
series occurring in YB are convergent for the considered range of values of the variable. In
particular, the geometry of the derivation of the ‘Gregory’ series works as it is described
only for angles in the first octant which, together with the fourth, fifth and eighth octants,
is also the range in which the series converges.

IX. THE GENEROSITY OF LANGUAGE, THE POWER OF ABSTRACTION

Reading YB from today’s vantage point, it is possible to imagine several directions along
which the development of calculus as depicted in it could have been done with greater
generality, without straining the abundant conceptual and computational resources it draws
upon. To try to identify these roads not taken and to speculate on why they were not taken
is, in spite of the risks, a temptation not easily resisted. I shall limit myself to one particular
instance of a missed turn which, in my view, clearly brings out the possible role of the
language in which the mathematics was ‘done’, as distinct from how it was communicated,
in inhibiting the requisite degree of abstraction and generalisation.
But first a remark, not directly related to language, on an issue YB does not concern
itself with, that of the irrationality of π. At first sight this is a surprising omission, given
the central position occupied by the geometry of the circle in YB’s calculus, especially if, as
I have suggested above, the infinite series expansions grew out of the search for a method
of ‘controlling’ an irrational number like π. Also as we have seen, the person who asserted
this irrationality was none other than Nilakanthan, Jyeshthadevan’s teacher and the author
of Tantrasam
. graha, the source book for YB. One reason for the silence may be technical,
that a yukti for the assertion was beyond the computational methods at the command of
the Kerala school – after all, it was only in 1761 that in Europe Johann Heinrich Lambert
proved Nilakanthan’s conjecture (without of course knowing it as such). But a more likely
explanation may lie in the philosophical underpinning of Indian science as a whole through-
out its long history. Indian mathematicians, unlike the Greeks, seem never to have come
√
to grips with proofs of the irrationality of any number, not of the far easier case of 2 nor
√
of 10, often used as an approximation for π at the time of Aryabhata. Almost certainly,
this failure has to do with the necessity of having to use reductio ad absurdum methods in
any such putative proof. For the Greeks, proof by contradiction, involving as it does the

21
principle of the excluded middle, was legitimised by the authority of Aristotle. But in India
the view that P and notP are complementary and mutually exclusive was explicitly rejected
at least by some and from at least the time of the Buddha[16]. Subsequently, reasoning by
contradiction (tarka) was much debated but generally not accorded full admissibility[17].
Like all scholars, mathematicians would have had a good grounding in other sciences as well
(this certainly was the case in Kerala at the relevant time[18]), and it is almost inconceiv-
able that they did not know of the reservations regarding recourse to tarka as a means of
establishing a logical truth – YB contains not a single instance of it; elsewhere in India, it
is not till a hundred years later that we come across its first hesitant and insecure use in
mathematics[19]. The fact remains that this disdain denied Indian mathematics the use of a
powerful proof device. Contrast this with Newton’s free use of it in the Principia (Book 1)
whose reliance on infinitesimal geometry (but with diagrams supplied), especially of similar
triangles, is otherwise very reminiscent of the Kerala techniques.
The missed turn I wish to focus on concerns the generalisation of Madhavan’s series
expansion of sin θ in powers of θ with numerical coefficients (‘around θ = 0’) to an expansion
around some nonzero value of θ, namely an expansion of sin(φ + θ) in powers of θ with
coefficients depending on φ. This would amount to a generalisation of the Maclaurin series
for the sine function (and the cosine function) to the corresponding Taylor series. Since such
expansions of general (sufficiently ‘good’) functions f :

df y 2 d2 f
f (x + y) = f (x) + y + + ...
dx 2! dx2
are relatively early landmarks in European calculus, the question whether Madhavan’s in-
terpolation formula can be thought of as giving the first few terms of the Taylor series of the
sine function has recently provoked some debate[20]. It may or may not be a satisfactory
interpolation, but it certainly cannot be thought of as the beginning of the Taylor expansion
– there is a numerical mismatch in the coefficient of the cubic term.
If I were time-transported to the 15th-16th century as a worthy member of the Kerala
community of mathematicians I could have pointed out (in YB’s irresistible prose style)
that, combining i) jīveparasparam (the formula for sin(φ + θ)), ii) jyānayanam (the method
of determining an arbitrary chord to arbitrary accuracy, i.e. Madhavan’s sine and cosine
series) and iii) bhujākhand
. am and bhujākhand
. āntaram (the first and second differentials of

22
sine and cosine), the (correct) Taylor series for the sine function will result. In detail,

sin(φ + θ) = sin φ cos θ + cos φ sin θ

θ2 θ3
= sin φ(1 − + ...) + cos φ(θ − + ...)
2! 3!
θ2 θ3
= sin φ + θ cos φ − sin φ − cos φ + ...
2! 3!
This is computationally trivial and, by truncation, produces an interpolation formula to any
order in θ. Furthermore, all derivatives of the sine function are determined by khand
. āntaram
(as also I would have pointed out):
d2n+1 sin φ d2n sin φ
= (−1)n cos φ, = (−1)n sin φ
dφ2n+1 dφ2n
and similarly of the cosine function. So Madhavan’s expansion, when shifted from θ = 0 to
some (any) other point on the circle, together with a realisation that, abstractly, a circle (a
‘uniformly round’ figure) does not distinguish one radius from another, is exactly the Taylor
series.
At issue here is not whether calculus was invented in India. That question is already
answered in the affirmative in the material available in YB: though in a very different
mathematical language, it is unmistakeably the calculus of Newton and Leibniz, but applied
1
only to functions corresponding to arcs of the circle, y = (1 − x2 ) 2 or, parametrically,
x = cos θ, y = sin θ. What Kerala mathematics did not have was an appreciation of the
great generality of the concepts and methods it had in hand and deployed so effectively;
what was missed is the power of abstraction.
To fully appreciate the extent of this power, we only need to call to mind how math-
ematics (and, to a lesser extent, the other exact sciences) has evolved over the past few
centuries, driven as it is by an increasing emphasis on the structural properties of the sets of
objects being studied, to the disadvantage of predominantly calculational techniques. When
mathematical objects belonging to a set are characterised by the operations that can be
carried out on them and the rules the operations have to obey (the relations amomg them)
we have, effectively, defined those objects in terms of the structure of the set. An object
has those properties and only those that come from its belonging to a structurally defined
set, bringing with it the freedom to transcend the particular circumstances in which it may
first have presented itself. Thus we may wish to study the set of transformations that take
a 3-dimensional cube into itself and may soon realise that the resultant of any two such

23
symmetry transformations is also a symmetry transformation, their product, uniquely de-
termined by the first two taken in order; that every symmetry can be undone by another,
its inverse, also uniquely determined; etc., in short that it belongs to a class of sets called
groups. Already at this stage we have gone through two levels of abstraction. At the first
level, we have written down the multiplication table – the specification of the resultant of
each ordered pair of symmetry transformations – for the group of symmetries of the cube.
With this table in hand, we know the group fully: there is no proposition concerning such
transformations, no matter how complicated or subtle, whose truth cannot be decided by
the table alone, without reference to the cube. It follows that all groups of transformations
of any ‘physical’ entity whatever, indeed all groups not even necessarily of transformations,
are abstractly the same group if they have the same multiplication table. At the second,
deeper, level, we would recognise that, though the group of symmetries of a cube and of, say,
a regular tetrahedron are not the same, i.e., do not have the same multiplication table, not
even the same number of elements, they still share a common abstract structure, the struc-
ture of a group. We would then proceed to elaborate the (essentially syntactic) properties
that all groups must have; in short we would make a general theory of groups.
It is obvious that I could have made my point more concisely and elegantly, in fact more
powerfully, if I had taken recourse to the language that developed to accommodate the
abstract structural point of view, a language primarily algebraic: symbols for the various
groups under consideration and for their elements, a notation for the group operations,
an economical statement of the relations, etc., the language in which theorems/‘truth’ will
be presented/expressed. Everyone now knows that group structures occur in virtually all
sciences, that modern physics in particular has been revolutionised by a systematic exploita-
tion of the formal, syntactic, understanding we have acquired of these structures. As for
mathematics, and staying with groups, I only mention the recently completed classification
of all finite simple groups (the recursive building blocks of all finite groups) as a general
illustration of the generative power inherent in the artificial language of algebra. Without
that as a vehicle of thought and communication, it seems inconceivable that we could have
reduced mathematical reasoning and creativity to an application of “rules without meaning”
so successfully.
Rules without meaning but not without purpose. Decisions about what abstract (syntac-
tic) structures are likely to prove ‘interesting’ and even judgements about what results are

24
‘deep’ or ‘beautiful’ have still to be made semantically, by reference to the contexts in which
questions arise and are resolved. The tension between problem-solving and theory-making,
the virtual inseparability of content and language, is arguably the main source of the ex-
tarordinarily wide scope and power of mathematics as it is practised now (and is perhaps
also the reason for the “unreasonable effectiveness of mathematics in the physical sciences”).
Once the creative process of discovery is thus algebraised and put on autopilot as it were,
there is no limit to what is discoverable – algebra is truly, infinitely, generous.
Moving back from these generalities to the Kerala of the 15th and 16th centuries, one
cannot of course be dogmatic about the extent to which progress towards greater abstraction
and hence greater generality was inhibited by the spurning of symbolic methods. But on the
evidence of YB, it cannot be doubted that obvious general points of view were overlooked
or neglected. As the episode of the Taylor series makes clear, an intuitive feeling for the
symmetry of the circle (the corresponding group being the group of all rotations around the
centre, which is not a finite group but still, structurally, a group) did not translate into the
precise understanding that this symmetry denies a privileged role to any particular diameter,
even if it is called the east-west direction. I like to think that if only they could have been
persuaded to designate a diameter as the line AB or αβ or a Malayalam equivalent thereof,
the Kerala mathematicians might have gone on eventually to a more general, less concrete,
appreciation of their own achievement.
What is puzzling in all this, at least to a non-expert, is that abstraction and symbolic
representations of objects and their relationships were not strangers to Indian thought.
Grammar (first and foremost), prosody and the classification of meters, cosmogonic specu-
lation, systematisation of the rules of reasoning and several other areas of intellectual inquiry
come to mind as having been greatly enriched by a fundamentally structural point of view.
In mathematics, the positional notation for numbers is itself a triumph of structure over
clumsy description; as we have seen, it is pure syntax. Bhaskara the second, who lived
not more than three centuries before the efflorescence of mathematics in Kerala and whose
work was known and cherished there, was a true algebraist; his Bījagan.ita proposes the use
of the names of colours to symbolically represent variables in equations, as distinct from
numerical values for them, and states the rules for manipulating them, exactly as if they
were numbers. (In a strange echo, modern physics employs the term chromodynamics for
the interactions of quarks of different ‘colours’). Bhaskara’s algebraic legacy seems to have

25
been better nourished in the Arabic lands than among his own mathematical heirs in India.
Perhaps, at the time and place we are talking about, mathematics was seen as no more than
a handmaiden of astronomy, with no high intellectual standing of its own. But then, what
else but intellectual curiosity can have inspired a Nilakanthan to cogitate on the irrationality
of π or a Madhavan to compute it to thirteen decimal places?
Close to three hundred years after Madhavan, Newton and Leibniz laid the foundations
of calculus as we know it, in the form and language that we follow, more or less faithfully,
today. That came very soon after the first integrations were performed in Europe, by Cav-
alieri, Fermat and a host of others (perhaps only coincidentally, the integral of x2 ). The
context in which this happened was vastly different from the one in which the heroes of
our narrative perfected their infinitesimal geometry of the circle. Descartes had already
algebraised geometry[21] and both Newton and Leibniz were well versed in the Cartesian
method and philosophy. The first ideas about a general function of a variable were taking
shape and the first correspondences between functions and the simplest curves, the conic
sections, beginning to be understood. The Newton-Leibniz calculus reflects this great open-
ing up, especially in the willingness and ability to handle more general functions. (In the
case of Newton, the Newton of the Principia in particular, a degree of generality could not
have been avoided; the position of a mass-particle as a function of time is a priori unknown,
being determined, in terms of a given force, only after the equations of motion are solved.
In contrast, with the early exception of Aryabhata with his belief in a spinning earth and
ideas on relative motion, Indian theories of the motion of celestial bodies were almost en-
tirely phenomenological (in the modern physicist’s usage of this term), paying little heed
to possible underlying causes). It is in this fundamental respect, in the recognition that
the infinitesimal method is of universal applicability, that the European calculus of the late
17th century – even while admitting that an acceptable foundation for it was not laid till the
middle of the 19th – can be seen to have gone far beyond Madhavan’s pathbreaking achieve-
ments. If history were different and granted them the time, could Madhavan’s followers
have taken their own road one day to this high ground? Going by what YB tells us, we
must remain sceptical. For one thing, there is the much-discussed difference in the Europian
and Indian approaches to doing mathematics, summarised in the catch words ‘deductive’
or ‘axiomatic’ and ‘constructive’ or ‘computational’. Our look at (a tiny sample of) YB’s
contents provides, I think, enough evidence that a more severe obstacle was the aversion to

26
a spare and refined language and the consequent absence of a structural or general point of
view.
But what history brought was the Portuguese incursion into Kerala, signalling another
period of strife and disorder. The resistance to the invaders was led by the Zamorin who, in
his other traditional role, was overlord and protector of the temples and patron of learning.
The delta of the river Nila saw much violence and bloodshed during the very time in which,
half a day’s walk away, Nilakanthan, Jyeshthadevan and others were teaching and writing
down their new mathematics in their temple villages. All the great texts date from this
turbulent 16th century and with its end mathematics in Kerala also went into terminal
decline.
Acknowledgements. Over the years of my involvement with YB I have frequently enjoyed
the generous hospitality of the Institute of Mathematical Sciences, Chennai, for which I
express my gratitude. My thanks go also to several friends, in particular Frits Staal and
David Mumford, for encouragement and much critical stimulation and advice.
Notes
[1] “On the Hindu quadrature of the circle and the infinite series of the proportion of the
circumference to the diameter exhibited in the four śāstras Tantrasamgraha, Yuktibhasha,
Caranapaddhati and Sadratnamala”, published in the Transactions of the Society 3(1835),
p.509.
[2] As far as I can tell, Whish’s baton was first picked up by K. Mukunda Marar and
C. T. Rajagopal, “On the Hindu quadrature of the circle”, J. Royal Asiatic Soc.(Bombay
branch) 20(1944), p.65. Of the numerous subsequent publications, two are worthy of special
mention: T. A. Sarasvati Amma’s thesis (Madras University) later published as Geometry
in Ancient and Medieval India (1979; second revised edition 1999), Motilal Banarsidass,
Delhi; and George Gheverghese Joseph, The Crest of the Peacock: Non-European Roots of
Mathematics (1992), Penguin Books, London. The first is largely faithful to the methods
actually employed by the Kerala mathematicians. Joseph’s book has brought the story to
a large readership. There are a fair number of good websites that discuss the work of the
Kerala school and there are also some which are not fully reliable.
[3] Rama Varma (Maru) Tampuran and A. R. Akhilesvara Ayyar, Yuktibhās.ā (Part I -
General Mathematics), with vyākhyā (1948), Mangalodayam Ltd., Trissivaperur.
[4] The question of why YB was composed in Malayalam is an intriguing one, not addressed

27
here.
[5] This extremely concise sketch of the medieval history of Kerala is an inadequate summary
of the pioneering studies of Ilamkulam (or Elamkulam) P. N. Kunjan Pilla. Some aspects
of his work have recently been questioned; despite this, his research provides the one coher-
ent overall picture we have of the history of Kerala, especially of the time relevant to us.
Ilamkulam wrote in Malayalam, but there is a volume of English translations of a selection
of his essays: Studies in Kerala History (1970), National Book Stall, Kottayam. He has the
added distinction of having been the first to seriously read Shankaranarayanan’s commen-
tary on the Laghubhāskarīyam, thereby establishing a first direct link between astronomical
activities in Kerala and elsewhere in India.
[6] I have chosen to transcribe all names as they are conventionally written in Malayalam.
Diacritical marks are supplied, generally, only for technical terms, also as they are written in
Malayalam script in YB (though they are of Sanskrit origin). I have ignored the Malayalam
distinction between the short and long e and o as well as the two l s.
[7] It is possible to make the case that Madhavan or his family was a recent arrival from
up north. He is said to have been an emprantiri (brahmins recently arrived from coastal
Karnataka, an established staging area on the migrant route). There is no known Sangam-
agramam (‘the village at the confluence’) in Kerala, if we set aside some fanciful etymology
of place names. Temples to Sangamesvara (Siva) stand at several confluences of rivers or
streams in northern Karnataka and southern Maharashtra.
[8] K. V. Sarma and S. Hariharan, “Yuktibhās.ā of Jyes.t.adeva”, Ind. J. Hist. of Sci.
26(2)(1991), p.185. The date suggested in this article is about 1530 CE but Sarma later
tended to favour a slightly later dating. Once it is accepted that YB is an account of
Madhavan’s mathematics, transmitted via Nilakanthan and others, a sharp date is of little
consequence except to historiographers.
[9] K. V. Sarma’s view ( “Aryabhata and Aryabhatan Astronomy: Antecedents, Status and
Development” in Proceedings of the International Seminar and Colloquium on 1500 Years
of Aryabhateeyam (2002), Kerala Sastra Sahitya Parishad, Kochi) that Jyeshthadevan also
authored Dr.kkaran.am, an astronomical chronicle datable to 1607, apparently on the basis
of hearsay gathered by Whish, is difficult to reconcile with the accepted date of YB and the
dramatically different linguistic styles of the two texts.
[10] An independent confirmation of Sarma’s view (see note [9]) will strengthen the case for

28
someone other than Jyeshthadevan, a disciple for example, doing the actual writing down
of the material. Edward Stadum, who was present at the Workshop, later used the phrase
‘literary calculus’ to describe YB. In my mind I can see Jyeshthadevan teaching under a
coconut tree, with a sign above saying ‘calculus spoken’.
[11] The Malayalam verb used is colluka which can mean state, relate, narrate, speak,
describe, explain, etc. and is employed in all these senses in YB.
[12] There are other roughly contemporaneous texts, though none in Malayalam, based on
Nilakanthan’s Tantrasam
. graha, for example Sankara Variyar’s Yuktidīpika. In many ways,
not least in the breadth of his interests, Nilakanthan appears to have been the real inheritor
of Madhavan’s mantle.
[13] Bhaskara II arrived at the correct formulae for the surface area and volume of the sphere,
but by seminumerical methods which the inventor of the infinitesimal method must surely
have scoffed at.
[14] Anterior to the Kerala work, sam
. kalita is the standard Indian term for the sum of any
(finite) series. According to B. Datta and A. N. Singh, “Use of series in India”, Ind. J.
Hist. of Sci. 28(2)(1993), p.103, this usage goes back at least to the (ambiguously dated)
Bakhshali manuscript.
[15] John Kadvany, “Positional Notation and Linguistic Recursion”, lecture at the Workshop.
[16] Scepticism about the validity and/or usefulness of the rule of the excluded middle was
expressed by the Buddha himself, according to the Pali canon. The famous list of questions
on which the Buddha declined to take a position includes some concerning the nature of
the physical world, e.g.: Does the universe have a finite extent or not or neither or both?
The great congress (3rd century BCE) which saw the parting of the Mahasanghikas from
the orthodoxy following a ferocious debate, and whose proceedings are recorded in the
Kathāvattu, not only made references to such ‘multi-valued’ propositions from the Buddha’s
teachings, but also discussed rules of debate and disputations.
[17] In Indian philosophy, the term tarka signifies “some sort of reductio ad absurdum where
an appeal to some absurdity or absurd consequence is made in order to lend an indirect
support to a positive thesis [by showing] in fact that the opposite thesis leads to absurdi-
ties” (Bimal Krishna Matilal, Perception: An Essay on Classical Indian Theories of Knowl-
edge(1986), Oxford University Press, New Delhi, p.79). The widespread reluctance to accept
tarka as “a means leading to a positive piece of knowledge” is also discussed here.

29
[18] For an examination of the epistemological roots of Kerala mathematics, from the writings
of its most thoughtful and articulate representative (Nilakanthan), see Roddam Narasimha,
“The Yukti of Classical Indian Astronomical Science”, to appear. It would seem that the
study of the impact of philosophical systems on mathematical thinking in the Indian context
is still in its infancy. I thank Frits Staal for a correspondence on the possible links between
tarka and reductio ad absurdum (or their absence!).
[19] M. D. Srinivas, “Proofs in Indian Mathematics” in Contributions to the History of
Indian Mathematics, G. Emch, R. Sridharan and M. D. Srinivas, Eds. (2005), Hindustan
Book Agency, New Delhi.
[20] A good starting point for following the debate regarding Madhavan’s ‘wrong’ interpo-
lation formula is Kim Plofker, “Relations between Approximations to the Sine in Kerala
Mathematics”, in Emch, Sridharan and Srinivas, cited above. The article is also a valu-
able step-by-step guide to the computation of the interpolation formula as given in Sankara
Variyar’s Yuktidīpika. YB itself does not mention the topic.
[21] Two and a half centuries before Descartes, Nicole Oresme (or Nicolas d’Oresme) had
taken the first step to liberate space from its Euclidean literalness and give it a more
metaphorical role by plotting on a plane the graph of a ‘function’ of a ‘variable’. I owe
my acquaintance with Oresme’s work to David Mumford, who is also responsible for the
characterisation above of its significance. The Indian view of geometry is, if anything, more
down-to-earth than the Greek – compare i)the use of the terms bhūmi and base for the
‘horizontal’ line in a polygon in the two traditions, ii)YB’s instruction to ‘drop a plumbline’
with drawing (or dropping!) a perpendicular, not necessarily to a base, and several other
similar instances. Notable coincidence: Oresme and Madhavan (if we accept the conven-
tional chronolgy) lived at about the same time, the former being the senior by twenty or
thirty years.

30